To use v0, go to v0.app, sign in, and describe the app or page you want in plain language — v0 generates working Next.js, Tailwind, and shadcn/ui code with a live preview, and you keep refining it by chatting until it's right, then publish straight to Vercel. The whole loop is prompt, preview, edit, deploy, with no separate build step in between.
Short answer: Sign up free at v0.app, start a chat, and describe what you want to build. v0 generates a working Next.js app you can preview instantly, tweak by chatting or with Design Mode, and publish to Vercel from the top-right Publish button. Free gives you $5 in monthly credits and 7 messages a day; paid plans start at $30/user/month.
I spent a few sessions building the same kind of thing most people actually try first — a small internal dashboard with a data table and a form — to see how far a single prompt gets you and where you actually have to start directing it. Below is the setup, the steps in the order I'd learn them, prompts that worked, and the mistakes that wasted my credits before I figured out why.
What you'll need
Nothing beyond a browser and a Vercel account, which doubles as your v0 login since both run on the same account system. No payment method is required to start — the free plan gives you $5 in credits a month and works without a card. If you want your generated app in version control, connect a GitHub account from the project menu; this isn't required to build or preview, only to push code out of v0. If you plan to deploy for real (not just preview), you'll eventually want a Vercel project to publish into, which v0 creates for you automatically the first time you hit Publish. Everything in this guide works on the free tier, credits permitting.
Step-by-step: How To Use V0
1. Sign in and start a new chat
Go to v0.app and sign in with your Vercel account (or create one). Click into the prompt box on the homepage — there's no separate "new project" step first; typing your first prompt and hitting enter starts both the chat and the build.
2. Describe what you want to build
Write a plain-language description of the app or component, the same way you'd brief a developer. "A task tracker with add, edit, and delete" and "a dashboard showing user analytics with charts" are the kind of prompts v0's own docs use as starting examples, and in my testing that level of specificity is close to the floor for a usable first result — vaguer than that and you mostly get a generic layout you'll spend more time correcting than you saved.
3. Review the generated preview
v0 returns working code and a live, clickable preview side by side, built on Next.js, Tailwind CSS, and shadcn/ui components. In my testing, a single well-specified prompt reliably produced a page that rendered without errors on the first pass — the gap was almost always in the details (wrong field types, missing empty states), not broken code.
4. Refine by chatting, one change at a time
Reply in the same chat thread instead of starting over: "add a search bar above the table" or "make the delete button ask for confirmation first." Each message is a real diff against your current version, not a regeneration from scratch, so earlier changes stick.
5. Switch to Design Mode for visual tweaks
Click the cursor icon in the chat input, or press Option+D (Alt+D on Windows), to enter Design Mode. This lets you click an element directly in the preview and adjust its spacing, color, or type instead of describing the change in words — faster for anything you can point at, slower than chat for anything structural like data flow or logic.
6. Drop into code view when you need precision
Switch to Code view to read or hand-edit the actual files v0 generated. I used this most for adjusting exact copy and fixing one prop name v0 had guessed wrong — small, surgical edits are faster to just type than to describe back to the chat.
7. Connect GitHub if you want the code out of v0
Open the project menu (the "…" icon) → GitHub, and create a repository. This pushes your current code into a real repo you can clone, branch, and review PRs against — useful the moment you want this to live alongside the rest of your codebase instead of only inside v0.
8. Publish to Vercel
Click Publish in the top right, then Publish to Production. Deployment took under a minute in my testing each time, and you get a live URL immediately. Further chat or Design Mode edits don't go live automatically — you click Publish Changes again when you're ready to push an update.
Example prompts you can copy
These are close to what I typed while testing the steps above:
- First prompt: "A customer feedback dashboard with a table of submissions, a status filter (new/reviewed/resolved), and a form to add a new entry manually."
- Follow-up refinement: "Add sorting to the table by date, and show the status as a colored badge instead of plain text."
- Data shape: "Replace the mock data with a Supabase table called
feedbackthat has columns: id, name, message, status, created_at." - Design Mode-adjacent, but chat-driven: "Increase the spacing between rows in the table and make the header row sticky when scrolling."
- Error handling: "Add a loading state while data fetches and an empty state when there are zero results, instead of a blank table."
Name the exact field, component, or behavior — "make it better" gets a generic pass, while "make the header row sticky" or naming your table's actual columns gets something you can ship with one more look.
Common mistakes to avoid
The mistake that burned the most credits early on was re-describing the entire app in one giant prompt after a small thing went wrong, instead of just naming the one thing to fix — v0 handles a targeted "the delete button doesn't ask for confirmation" far better than a restated wall of requirements, and it costs less. Second, skipping Design Mode and describing purely visual changes in words ("make the button a bit more to the left") when clicking the element and dragging would take five seconds. Third, not connecting GitHub until much later, which meant losing the easy diff history I'd have had from the start. Fourth, assuming the free plan's 7 messages a day resets faster than it does — I hit the cap mid-session more than once and had to wait for the next day. Fifth, publishing straight from a first-draft chat without opening Code view to sanity-check anything touching real data — the preview looking right and the underlying query being right aren't always the same thing.
v0 plans compared
| Plan | Price | Included credits | Best for |
|---|---|---|---|
| Free | $0 | $5/month, 7 messages/day cap | Trying v0, small one-off components |
| Plus | $30/user/mo | $30/month + $2 daily login bonus | Regular solo or small-team building |
| Business | $100/user/mo | $30/month + $2 daily login bonus, training opt-out | Teams needing centralized billing and data controls |
| Enterprise | Custom | Custom | Orgs needing SSO, RBAC, and support SLAs |
Beyond the subscription, v0 also bills model usage separately by the token if you exceed your included credits — v0 Mini runs $1/$5 per million input/output tokens, up through v0 Max Fast at $10/$50, per v0’s pricing page, checked here on August 4, 2026. Every plan gets you the same Next.js/Tailwind/shadcn output; what changes is how much you can generate before you're paying per token.
Tools that make this easier
v0 is one option in a crowded field of AI coding tools, and where it fits depends on what you're building. If you want a traditional editor with AI woven in rather than a chat-to-app builder, my how to use Cursor AI guide covers the closest alternative for people who want to work inside a full codebase, not just a generated one. For a head-to-head on agent-style coding assistants, Cursor vs Copilot tests both on the same brief. If you're deciding between v0 and a general-purpose chatbot for coding help, ChatGPT alternatives for coding rounds up the field with real prices. Claude powers a lot of AI coding tools under the hood, so how to use Claude AI is useful background if you're curious what's actually generating v0's suggestions on the backend for some models. For the broader picture of what "best" even means in this category, best AI tool for code compares v0-style builders against IDE-native tools directly. And if you're still deciding whether an AI coding assistant fits your workflow at all, AI coding assistant is the right starting point.
My take
v0 is genuinely fast for the thing it's built for: getting from a plain-language idea to a working, deployable Next.js app without opening an IDE first. It's not a replacement for a real editor once the project gets complex — I still dropped into Code view constantly for precise fixes, and Design Mode has real limits once you're past surface-level styling. The free plan's 7 messages a day is enough to evaluate it seriously but not enough for daily use; Plus at $30/user/month is where this becomes a real tool rather than a demo. If your starting point is usually "I need a working prototype fast," v0 earns a spot in the rotation. If you're maintaining a large existing codebase, an editor-native tool like Cursor will likely serve you better day to day.
Frequently Asked Questions
How To Use V0: is it free?
Yes, to start. The free plan includes $5 in credits per month and a cap of 7 messages a day, enough to build and preview a small app. Regular use runs into both limits quickly, and the next tier, Plus, starts at $30 per user per month.
How long does it take to how to use v0?
A first working preview typically takes one prompt and under a minute. Getting a result you'd actually publish usually takes several rounds of chat-based refinement plus a pass in Code view — in my testing, a small dashboard-style app took about 20-30 minutes from first prompt to something I'd deploy.
What is the easiest way to do this?
Write your first prompt like a brief, not a topic — name the specific features and data involved, the way you'd explain the app to a developer. Then fix one thing at a time in follow-up messages rather than re-describing the whole app when something's off.
Do I need to know how to code to use v0?
No, for a first pass — you can build, preview, and publish an app through chat and Design Mode alone. But if anything breaks or needs a precise fix, reading the generated code in Code view helps a lot, so some familiarity with reading code is genuinely useful, not just optional.
Can I export the code v0 generates?
Yes. Connect GitHub from the project menu to push a real repository, or copy the code directly from Code view. The output is standard Next.js, Tailwind CSS, and shadcn/ui — nothing proprietary that locks you into v0 once you've generated it.